AI4OPT Seminar Series

Date: Thursday, August 13th, 2026

Time: Noon – 1:00 pm

Location: Coda (756 W Peachtree St NW, Atlanta, GA 30308) - (Druid Hills Conference Room C1115)

Speaker: Xinliang Dai

Two Scalable Distributed Strategies for Nonlinear Optimization in Power Systems

Abstract: Large-scale power-system optimization combines sparse network physics, many local states, low-dimensional interfaces, and distributed data ownership. This presentation introduces two complementary strategies for nonconvex nonlinear programming: condensing local information during the solve via iterative distributed optimization, or before the solve via non-iterative flexibility aggregation.

The first, Barrier ALADIN (BALADIN), retains the full regional nonlinear models and coordinates parallel local barrier subproblems through Schur-complement condensation of local KKT systems. On large-scale AC optimal power flow benchmarks, it solves systems with up to 193,000 buses and achieves 2-5X speedups over centralized IPOPT. The second strategy replaces each detailed distribution-system model with a compact predictor-corrector surrogate of its implicit feasible set in low-dimensional coupling variables. The surrogate has local cubic-order accuracy and can be constructed independently across subsystems and time periods. Across 31 radial and meshed networks, maximum sampled false- and lost-flexibility rates are at most 0.82% and 0.27%, respectively, while 24-period integrated transmission-distribution coordination achieves 6X end-to-end speedups. Together, the methods show that interface density is central to scalable distributed optimization in energy systems.

Bio: Xinliang Dai received the B.Sc. degree from Jilin University, China, and the M.Sc. and Ph.D. degrees from the Karlsruhe Institute of Technology (KIT), Germany. He is currently a Postdoctoral Research Associate with the Zero-carbon Energy Systems Research and Optimization Laboratory (ZERO Lab) at Princeton University, USA. 

His research focuses on graph-based distributed optimization, sensitivity analysis for energy systems, and GPU acceleration for large-scale optimization.

Note: Snacks will be served at the seminar. So, please stop by 15 minutes beforehand to serve up and be seated on time.

To continue receiving all AI4OPT seminar announcements, please sign up to the mailing list at https://lists.isye.gatech.edu/mailman/listinfo/ai4opt-seminars
Videos of the past seminars can be seen on AI4OPT webpage at https://www.ai4opt.org/seminars/past-seminars